A surge of interest in an artificial intelligence chatbot developed by Chinese startup DeepSeek rattled markets and reignited debates about economic and geopolitical rivalry between the United States and China over advanced AI technology.
DeepSeek’s AI assistant briefly became the most downloaded free app on Apple’s App Store, driven by widespread curiosity about a potential competitor to ChatGPT. Observers in the U.S. tech sector were unsettled by claims that the Chinese startup had produced models that match the capabilities of leading American systems while requiring far less computing cost.
If those cost claims hold up, they could challenge the assumptions behind the massive infrastructure spending U.S. tech firms plan for data centers and high-end chips to support further generative AI development. At the same time, a great deal of the reaction was shaped by hype and misunderstanding about what DeepSeek had actually achieved.
“Their models are impressive, but they aren’t miraculous,” said Bernstein analyst Stacy Rasgon, who covers the semiconductor industry. Rasgon and other market analysts described Wall Street’s early response as overblown. “They aren’t using any secret innovations; these are techniques others are also experimenting with,” he added.
What is DeepSeek?
DeepSeek, founded in 2023 in Hangzhou, China, released its first large language model that same year. The company’s CEO, Liang Wenfeng, previously co-founded High-Flyer, a prominent Chinese hedge fund known for AI-driven quantitative trading.
The startup drew wider attention last month after unveiling a new model it claimed could rival systems from U.S. companies like OpenAI’s ChatGPT while being more economical in its use of costly Nvidia processors during training. The chatbot became broadly accessible after appearing in major app stores earlier this year.
The most intense reaction followed a research paper published last week — coinciding with the inauguration of President Donald Trump — describing a DeepSeek model called R1. The paper highlighted advanced reasoning abilities, such as the model’s capacity to revise its approach to solve a math problem, and presented cost comparisons suggesting R1 was substantially cheaper to train than a comparable OpenAI model known internally as o1.
“I don’t know the exact economics, but the suggested price points alarmed many people,” Rasgon said. Much of the market turbulence stemmed from interpreting those numbers as definitive proof that DeepSeek had leapfrogged U.S. approaches to model design and efficiency.
What DeepSeek means for U.S. companies
Beyond technical claims, DeepSeek’s emergence has intensified a policy and strategic debate in the United States about how best to respond to competition from China in AI.
Venture capitalist Marc Andreessen described the development as “AI’s Sputnik moment,” invoking the 1957 Soviet satellite launch that ignited Cold War concern and a race in space technology. Andreessen and others argue that heavy-handed regulation of AI could handicap American innovation and inadvertently advantage Chinese firms.
At the same time, the DeepSeek story complicates U.S. efforts to slow China’s access to advanced AI hardware. For several years Washington has restricted sales of U.S.-designed AI semiconductors to Chinese buyers as a way to limit Beijing’s capacity to build state-of-the-art models. Some analysts view the timing and publicity of DeepSeek’s release as politically motivated.
“The underlying technological progress appears real, but the timing of the release has a political aspect,” said Gregory Allen, director of the Wadhwani AI Center at the Center for Strategic and International Studies. He compared DeepSeek’s announcement to prior instances in which Chinese companies used product launches to make a point during diplomatic disputes.
Supporters of export controls say the measures remain important to slow the transfer of specialized compute capability that accelerates model training. Critics warn that overly strict controls could spur alternative approaches that achieve similar outcomes without the same hardware, and that geopolitical signaling may be as influential as technical reality in shaping perceptions.
On his first day in office, President Trump signed an order directing his administration to “identify and eliminate loopholes in existing export controls,” a sign he intends to continue—and possibly tighten—the policies adopted under the previous administration.
Read more about tech stocks:
- How Nvidia became a strong investment: a look at the numbers
- Making sense of the markets: perspectives on 2025
- AI-focused ETFs in Canada: ways investors can participate in the AI trend
- Nvidia’s latest market valuation milestones